But then again, if we don't have a blueprint for superhuman AI, discussing how to make them safe is like discussing the sex of angels.
Or rather, like discussing how to make turbojets safe and reliable circa 1920.
Turbojets are unbelievably safe and reliable.
It was a difficult
@ylecun
-
Superhuman AI Safety: Blueprint and Historical Parallels
By
–
-
AI Safety Concerns: Good Intentions Don’t Guarantee Correct Outcomes
By
–
There certainly are AI doomers who are just grifters seeking attention.
But there is no doubt that Geoff and Yoshua's intentions are noble.
One can be well intentioned and wrong. -
Piaget and Chomsky: Learning and Language Theory Debate
By
–
Oui. En fait, je me suis intéressé à l'apprentissage en lisant le débat entre Piaget et Chomsky "théories du language, théories de l'apprentissage" en 1980.
-
Should AI Weight Known False Evidence in Reasoning?
By
–
Do you give 10% weight to a proof that you know is wrong?
-
Collective Defense Against Existential AI Risks
By
–
The answer is no, with any budget.
And even if it were possible for someone to do this, it would also be possible for a bunch of people to design a system to take down the world ender. -
Chess AI Evolution: From Alpha-Beta to Monte-Carlo Tree Search
By
–
Chess players are all about planning.
It used to be Alpha-Beta tree search with a fixed, handcrafted evaluation function.
Nowadays, it's Monte-Carlo Tree Search with a policy network to pick good candidate branches and a Value network that learns to evaluate nodes.
But planning -
AGI debate video premieres in 75 minutes
By
–
The cideo of our AGI debate premieres 75 minutes from now.
-
Public Fear and Government Response to AI Risks
By
–
The public is being scared by doomers relayed by some media.
Understandably, they look towards the government to protect them against those imaginary risks. -
Self-Supervised Learning for Vision and World Models
By
–
Self-Supervised learning.
Particularly for vision and world models.